Re: revisit: bare carriage return in status.m4 / configure

[Steven R. Loomis]

Eric Blake wrote:
> Steven R. Loomis <srl <at> icu-project.org> writes:
>   
> > That works only on ASCII based platforms, and it requires tr (which 
> > wasn't formerly required, according to the thread quoted below).
> >     
>
> Autoconf is already a significant client of tr (for example, look at the output 
> of AC_CHECK_HEADERS).  We need to submit a bug report against standards.texi 
> that adds tr to the list of approved programs.  But using tr in config.status 
> is not an issue.
>   
That's good to know.  Sorry about the top-posting.
> The problem that config.status has to work around is that not all awk and sed 
> recognize the \r escape sequence; for those that don't, the use of a literal ^M 
> in the configure/config.status was deemed workable at the moment.  On the other 
> hand, our use of raw ^M is ASCII-agnostic:
>
> ac_cr='
> '
> ac_cs_awk_cr=`$AWK 'BEGIN { print "a\rb" }' </dev/null 2>/dev/null`
> if test "$ac_cs_awk_cr" = "a${ac_cr}b"; then
>   ac_cs_awk_cr='\\r'
> else
>   ac_cs_awk_cr=$ac_cr
> fi
>
> If we change $ac_cr to be \015, then $ac_cs_awk_cr will likewise be \015 in the 
> case where the \r escape sequence didn't work (and if it did work, then we know 
> \r and \015 are equivalent).  So, whether or not \015 is carriage return in the 
> current encoding, we are doing an idempotent replace.
>
> I agree that using raw ^M in a portable script comes with its own set of 
> portability problems, therfore I will commit a patch that switches over to tr 
> in the near future.  Thanks Steven for the report, and Thomas for the patch 
> idea.
>   
You are welcome, and thanks for looking into this.

Note previous similar suggestion here: 
http://lists.gnu.org/archive/html/autoconf-patches/2008-04/msg00007.html

-steven